Word Images Research Articles

Utilization of information from CNN feature maps for offline word-level writer identification

This paper proposes a text-independent method for authorship identification using handwritten word images. Our method is text-independent and imposes no limitations on the size of the input word images being analyzed. To begin with, the SIFT algorithm is utilized to extract key-point regions (fragments) across different levels of abstraction, encompassing allographs, characters, or character combinations. These fragments are subsequently processed through a CNN network, yielding feature maps corresponding to convolution layers. The information in these maps is then transformed into a fixed-dimension representation using a modified version of the HOG feature descriptor.The noteworthy contribution of our proposal lies in harnessing additional cues from CNN’s feature maps for writer identification. We propose a measure to gauge the importance or ‘saliency’ of feature maps within a CNN layer during training. This measure originates from applying Sparse Principal Component Analysis (SPCA) to Histogram Of Gradient (HOG) features. Once saliency values are obtained, we combine them with HOG representations to create writer descriptors tailored to a CNN layer. The derived descriptors are then passed through a set of SVM classifiers that are scored at two levels to determine the identity of the handwritten word image.The efficacy of our system has been demonstrated on the word-level data of the CVL, IAM and CERUG-EN datasets with an accuracy of 82.10%, 87.68%, and 75.80% respectively. The results obtained are shown to be promising when compared with previous works.

The current status and prospects of transformer in multimodality

At present, the attention mechanism represented by transformer has greatly promoted the development of natural language processing (NLP) and image processing (CV). However, in the multimodal field, the application of attention mechanism still mainly focuses on extracting the features of different types of data, and then fusing these features (such as text and image). With the increasing scale of the model and the instability of the Internet data, feature fusion has been difficult to solve the growing variety of multimodal problems for us, and the multimodal field has always lacked a model that can uniformly handle all types of data. In this paper, we first take the CV and NLP fields as examples to review various derived models of transformer. Then, based on the mechanism of word embedding and image embedding, we discuss how embedding with different granularity is handled uniformly under the attention mechanism in multimodal scenes. Further, we reveal that this mechanism will not only be limited to CV and NLP, but the real unified model will be able to handle tasks across data types through pre-training and fine tuning. Finally, on the specific implementation of the unified model, this paper lists several cases, and analyzes the valuable research directions in related fields.

‘The Breath of Every Living Thing’: Zoocephali and the Language of Difference on the Medieval Hebrew Page

Abstract The most remarkable feature of the Hammelburg Mahzor, a fourteenth-century German High Holiday book, is the inclusion of zoocephalic figures: humans with beastly heads. The purpose of this essay is to explore the semiotics and phenomenology of this specifically Jewish visual idiom, and to suggest that its presence lies at the intersection of language, philosophy, poetry, and history. In the Mahzor, zoocephaly signals distinction that collapses temporalities, tests the limits of alterity, and engages in a sophisticated word–image play that strives to establish visceral connections with the community of the manuscript's users. Hammelburg zoocephali invoke the fragility of the human condition by establishing reverberating relationships between themselves and other inhabitants of the Mahzor's pages: echoes of many, avatars of none. Outwardly monstrous yet emphatically human, these zoocephali prove to be particularly excellent images to think with about the place of Hebrew manuscripts in the long history of medieval visual culture.

Harnessing CRNN Technology for Image-Based Sequence Recognition

We are studying the challenge of finding text in everyday images using advanced computer techniques, specifically deep learning. Our system can accurately spot and read text in images. It uses Convolutional Recurrent Neural Networks (CRNNs), a type of deep learning model, to locate and understand the text. Instead of relying on manual rules, our model learns from the data to identify text patterns. Our projects unique feature is that our model is trained on single- word images, making it versatile in recognizing different text styles and backgrounds. We trained our model using a large dataset, including the “MJ Synthetic Word Dataset”, which consists of 50,000 out of which we have taken 35000 astraining images, 15000 as validation images. The results have been promising, indicating potential real-world applications such as searching for words in large sets of images. Key Words: Advanced computer techniques, Deep learning, System, Spot and read text, Convolutional Neural Networks (CRNNs), Deep learning model

The Experimental Study of the Image of Russian Color Term Blue in the Consciousness of Russian Culture Bearers

The article is devoted to the study of the color term blue. This color occupies an important place in the color picture of the world of representatives of Russian culture, which determines the relevance of the empirical study of the image of this color term from the point of view of psycholinguistics. The theoretical basis of the study was the work of A.A. Zalevskaya, A.P. Vasilevich, E.I. Goroshko.When organizing and conducting the research, we assumed that individual and cultural knowledge about color are united in the linguistic consciousness. So, any image in the consciousness of an individual has its cultural and personal component. The main methods used in the work are associative experiment, semantic differential, factorial and definitional analyses. We used the data from related sciences in the process of analyzing the results of the study. The study was conducted on the basis of Southwest State University among full-time and part-time students.The work shows the structure of the associative field of the color term blue in different time periods, which al-lows us to clearly demonstrate the dynamics of changes in this field over the past thirty years.The article also examines the gender structure of the associative field of the color term blue, highlights general gender-neutral associations, as well as associations that are typically female or typically male. In addition, a factor analysis of the results of the semantic differential is carried out from the point of view of the gender characteristics of the perception of color term blue.The methodology of organizing and conducting experimental research described in the paper can be success-fully used for psycholinguistic research of images of various stimulus words. The data obtained during the experiment can be used not only in linguistics, but also in marketing, psychology and advertising.

Bagging: An Ensemble Approach for Recognition of Handwritten Place Names in Gurumukhi Script

In this article, the authors present an effort to recognize handwritten Gurumukhi place names for use in postal automation. Five feature extraction techniques (zoning, horizontal peak extent, vertical peak extent, diagonal, and centroid) have been analyzed and optimized using Principal Component Analysis (PCA). Four classification methods ( k -Nearest Neighbor ( k -NN), decision tree, random forest, and Convolutional Neural Network (CNN)) have been utilized to classify the handwritten word images. To enhance the recognition results, the authors have employed Bootstrap Aggregation (Bagging) with a majority voting scheme. The authors used a public benchmark dataset of 40,000 handwritten place-name samples in the Punjabi language for their experimental work. The experiments were conducted using a 70:30 partitioning approach, where 70% of the data was utilized for training and the remaining 30% for testing. The system achieved a maximum recognition accuracy of 96.98% by utilizing a combination of zoning, vertical peak extent, and diagonal features, and a minimum Mean Squared Error (MSE) of 0.86% based on a combination of zoning and horizontal peak extent features with a majority voting scheme through ensemble (Bagging) methodology.

Penerapan Algoritma C4.5 Untuk Memprediksi Blog Yang Memiliki Peluang Juara

Blogs are currently not only used as online diaries or online journals. Blog content that supports text, images, videos, gifs, animations, pdfs and YouTube makes the use of blogs more than just online dairy. Currently blog is one of the media used by companies and organizations in promoting goods and services. The existence of bloggers is needed by companies and organizations to review their products and services with the aim of attracting online customers. One of the efforts to invite bloggers to be able to review products in an interesting way has led business actors to become interested in holding blog contests. In this research, a model was built to predict blogs that have a chance to win by applying the C4.5 decision tree algorithm. The prediction model was created using 100 blog contest participant data obtained from three blog competitions organized by ASUS which were obtained through the contest participant links found on the committee blog. From the processed dataset, supporting seven variables including Word Count, DA, PA, Image Template, Domain and Champion. The resulting prediction model is a decision tree with 7 attributes which produces 11 leaves and 18 trees with an accuracy of 74% with a precision of 0.735 and a recall of 0.740.

TextStyleBrush: Transfer of Text Aesthetics From a Single Example.

We present a novel approach for disentangling the content of a text image from all aspects of its appearance. The appearance representation we derive can then be applied to new content, for one-shot transfer of the source style to new content. We learn this disentanglement in a self-supervised manner. Our method processes entire word boxes, without requiring segmentation of text from background, per-character processing, or making assumptions on string lengths. We show results in different text domains which were previously handled by specialized methods, e.g., scene text, handwritten text. To these ends, we make a number of technical contributions: (1) We disentangle the style and content of a textual image into a non-parametric, fixed-dimensional vector. (2) We propose a novel approach inspired by StyleGAN but conditioned over the example style at different resolution and content. (3) We present novel self-supervised training criteria which preserve both source style and target content using a pre-trained font classifier and text recognizer. Finally, (4) we also introduce Imgur5K, a new challenging dataset for handwritten word images. We offer numerous qualitative photo-realistic results of our method. We further show that our method surpasses previous work in quantitative tests on scene text and handwriting datasets, as well as in a user study.

THE EDUCATIONAL AND UPBRINGING FUNCTIONS OF CHILDREN'S MUSICAL THEATRE IN THE INCLUSIVE EDUCATION ENVIRONMENT

The initial hypothesis of this article is to determine the educational and upbringing functions of children's musical theatre in the conditions of inclusive education as a triad: supervisor-director, participants-performers, and parent audience. Each of these links in the triad creates an expressive subsystem, and its comprehensive complex is subordinated to the sphere of the educational-artistic context of the emergence and implementation of both literary and musical creative concepts since the multifaceted problem of synthesizing arts has been and remains an actual problem of scientific research. A detailed description of the musical sound palette confirmed the validity of our research regarding the objective realization of musical works in performance as a procedural process in the formation of art images of sounds and words, the diversity of their perception of meaning by performers, which is an extraordinary factor in the formation of stage images in a play involving children and adolescents with special needs. Taking this into account, the director should rely on his entire previous background, worldview, general cultural, aesthetic, and pedagogical experience to develop the creative activity of children through theatrical activities, considering their psychological and physical peculiarities.
 The purpose of the article is to actualize the role of children's musical theatre as a genre-forming extracurricular institution in the conditions of inclusive education, to determine its educational and upbringing functions, to study the peculiarities of stage action in the artistic reproduction of images, to reveal the significance of the musical palette as an expression of the artistic idea for creating a creative atmosphere of the performance based on artistic-pedagogical experience, and to determine the cognitive needs of children and adolescents with special developmental needs through theatrical activity.

Some properties of morphic images of (eventually) dendric words

Some properties of morphic images of (eventually) dendric words

Evaluating synthetic pre-Training for handwriting processing tasks

In this work, we explore massive pre-training on synthetic word images for enhancing the performance on four benchmark downstream handwriting analysis tasks. To this end, we build a large synthetic dataset of word images rendered in several handwriting fonts, which offers a complete supervision signal. We use it to train a simple convolutional neural network (ConvNet) with a fully supervised objective. The vector representations of the images obtained from the pre-trained ConvNet can then be considered as encodings of the handwriting style. We exploit such representations for Writer Retrieval, Writer Identification, Writer Verification, and Writer Classification and demonstrate that our pre-training strategy allows extracting rich representations of the writers’ style that enable the aforementioned tasks with competitive results with respect to task-specific State-of-the-Art approaches.

Statistics and psychometrics for the CAT-N: Documenting the Comprehensive Aphasia Test for Norwegian

ABSTRACT Background Ivanova and Hallowell 2013 emphasise the importance of reporting on test development and psychometric properties of tests in international journals. Such documentation may serve as references for other test developers and enable researchers and clinicians to assess reliability and validity issues in tests made for a language unknown to them. The CAT (Comprehensive Aphasia Test) is a general aphasia test which examines linguistic skills broadly, within the cognitive neuropsychological tradition; it has been and is being adapted to a number of languages. Aims The aim of this article is to document the statistical procedures used in the development and standardisation of the Norwegian adaptation of the CAT (CAT-N), to document its psychometric properties, and to discuss validity and reliability issues. Methods & procedures The adaptation of the CAT-N involved careful design of subtests and test items, taking into account features like word frequency, imageability and phonological and other language-specific linguistic variables. The prototype was tested on a normative sample of 85 persons with aphasia and a control group of 84 persons without aphasia. The items of some subtests were reordered based on the norming. A new scoring scheme was developed for two subtests of Picture description. The CAT-N includes the Aphasia Impact Questionnaire (AIQ), which is a new patient reported outcome measure developed for the CAT. Outcomes & Results Statistical methods are documented and discussed. Descriptive statistics for subtests and linguistic domains are presented. Internal consistency and partial inter-rater and intra-rater reliability aspects are investigated and documented. Construct validity is investigated and documented by factor analysis. Sensitivity and specificity are investigated through pairwise comparisons for subtests and domains and the use of normal-language cutoff values. Concurrent validity is investigated through comparisons with results from an existing aphasia test for Norwegian (NBAA). Conclusions The CAT-N is shown to have good reliability and validity, and it distinguishes well between persons with and without aphasia. The article provides explicit documentation of design decisions which may be useful in future adaptations of the CAT.

Convolutional Neural Networks Trained to Identify Words Provide a Surprisingly Good Account of Visual Form Priming Effects

A wide variety of orthographic coding schemes and models of visual word identification have been developed to account for masked priming data that provide a measure of orthographic similarity between letter strings. These models tend to include hand-coded orthographic representations with single unit coding for specific forms of knowledge (e.g., units coding for a letter in a given position). Here we assess how well a range of these coding schemes and models account for the pattern of form priming effects taken from the Form Priming Project and compare these findings to results observed with 11 standard deep neural network models (DNNs) developed in computer science. We find that deep convolutional networks (CNNs) perform as well or better than the coding schemes and word recognition models, whereas transformer networks did less well. The success of CNNs is remarkable as their architectures were not developed to support word recognition (they were designed to perform well on object recognition), they classify pixel images of words (rather than artificial encodings of letter strings), and their training was highly simplified (not respecting many key aspects of human experience). In addition to these form priming effects, we find that the DNNs can account for visual similarity effects on priming that are beyond all current psychological models of priming. The findings add to the recent work of Hannagan et al. (2021) and suggest that CNNs should be given more attention in psychology as models of human visual word recognition.

Flotsam and Jetsam: Art, Allegory, and Shipwreck in the Twenty-First Century (II)

Abstract This allegorical postcard is organized around two groups of photographs. The first group was the result of a joint collaboration with the Vancouver artist Scott Saunders and produced photographs which have peppered several of my previous texts published by American, British and Canadian Studies. The second is a series of photographs taken by Scott Saunders from the window of his apartment in Vancouver in which he documents the street life constantly ebbing and flowing on the sidewalk below. The catalyst for bringing these two groups together was a photograph I took several years ago in Sambro, Nova Scotia (a small fishing village located just outside of the city of Halifax) depicting a forlorn sunken fishing vessel. The term “flotsam” is applied, according to the Oxford Reference Dictionary, to “the wreckage of a ship or its cargo floating on or washed up by the sea,” while “jetsam” describes the things or objects deliberately “thrown away, especially from a ship at sea and that float toward land.” Combined, these images of words and devastated human beings are caught in an apparently endless circulation of violence and contingency located at the heart of the urban fabric of a modernity bereft of any horizon of hope, redemption, or rescue.

Assessment of verb and sentence processing deficits in stroke-induced aphasia: the Italian version of the Northwestern Assessment of Verbs and Sentences (NAVS-I)

ABSTRACT Background The Northwestern Assessment of Verbs and Sentences (NAVS) assesses verb and sentence production and comprehension in aphasia. Results from the original English version and from its adaptation to German have shown that the NAVS is able to capture effects of verb-argument structure (VAS) complexity (i.e., lower accuracy for two- and three-argument vs. one-argument verbs) and syntactic complexity (i.e., lower accuracy for non-canonical vs. canonical sentences) in both agrammatic participants and individuals with fluent forms of aphasia. The NAVS has been recently adapted to Italian (NAVS-I) and tested on a group of healthy participants, with results showing longer reaction times to complex vs. simple verbs and sentences. Aims The present study aimed to test the ability of NAVS-I to i) capture verb/sentence production and comprehension deficits in Italian-speaking individuals with agrammatism or with fluent aphasia, and ii) differentiate individuals with aphasia from healthy age-matched participants, with the overall goal to validate its use in clinical practice. Methods & Procedures Forty-four healthy participants and 28 individuals with aphasia (10 with agrammatic speech production) were administered the NAVS-I, which includes tasks assessing production and comprehension of verbs requiring one, two or three arguments, as well as production and comprehension of canonical and non-canonical sentences. Outcomes & Results On the Verb Naming Task (VNT), better production of one- (vs. two- and three-) argument verbs was found in the agrammatic group, whereas, verb production in the fluent group was solely predicted by word length and imageability. No effects of argument optionality (i.e., greater difficulty for optionally transitive verbs than for 1-argument verbs) were found. Sentence-level tasks found no differences between the agrammatic and the fluent group in production or comprehension of both canonical and non-canonical sentences; rather, sentence comprehension accuracy was predicted by demographic variables and by aphasia severity. At the individual level, performance on the NAVS-I was significantly different from that of healthy speakers in 26/28 patients. Conclusions Data show that the NAVS-I is able to capture effects of argument structure complexity in verb production, and effects of syntactic complexity in sentence production and comprehension. In addition, our results point to verb production as the task with greater capability to differentiate agrammatism from other (fluent) forms of aphasia. The study provides support for the use of the NAVS-I in the diagnosis of aphasia, as it is able to detect language deficits at the individual level, even in participants with mild (residual) forms of aphasia.

Improving OCR Accuracy for Kazakh Handwriting Recognition Using GAN Models

This paper aims to increase the accuracy of Kazakh handwriting text recognition (KHTR) using the generative adversarial network (GAN), where a handwriting word image generator and an image quality discriminator are constructed. In order to obtain a high-quality image of handwritten text, the multiple losses are intended to encourage the generator to learn the structural properties of the texts. In this case, the quality discriminator is trained on the basis of the relativistic loss function. Based on the proposed structure, the resulting document images not only preserve texture details but also generate different writer styles, which provides better OCR performance in public databases. With a self-created dataset, images of different types of handwriting styles were obtained, which will be used when training the network. The proposed approach allows for a character error rate (CER) of 11.15% and a word error rate (WER) of 25.65%.

Philosophical and Legal Understanding of the Concepts: Image, Icon, Image, Honesty, Reputation, Their Manifestations in Criminal Procedural Law

Introduction. The article presents an attempt for the first time, on the basis of philosophical and legal understanding, to use the concept of “reputation”, which is new for legal scientific circulation. Legal terms, including criminal procedural ones, arise as generic terms. That is, concepts are brought under a common genus, and then specific differences are distinguished. Philosophical categories, due to their universality, cannot be summed up under the genus, and, as a rule, act as a dialectical pair of opposites the thesis and antithesis. Such, for example, are the concepts of essence and phenomenon, necessity and chance, content and form. The complexity of introducing a new concept of reputation into scientific circulation lies in the fact that it does not have its own dialectical antithesis. A bad reputation is also a reputation. At a stretch, reputation can be opposed to discredit, but discredit is an action aimed at undermining reputation, and this is a transition to another kind. Concepts of this kind can be meaningful by analogy. As such, the authors chose the dialectical pair icon image, like foreign-language tracing papers from the Russian word image, which in the legal sense are reduced to reputation. Both that and another (both an icon and an image) are an image, but how different and even opposite are the meanings of these words. Moreover, initially the intentions of the authors were precisely linguistic (and not religious or marketing), and only then the words themselves designated the cultural area of their application.Materials and methods. The theoretical and methodological basis of the study was the hermeneutical approach to the analysis of the concept of "reputation". Methods of systematic, comparative legal analysis, special philosophical and legal research tools were used. The authors used examples from various areas of law criminal procedure, criminal, civil.Research results. Based on the results of the analysis, the question of trust is raised in relation to the procedural status of participants in criminal proceedings. The introduction of the concept of "reputation" as a legal category into legal circulation is proposed.Discussion and conclusion. In the form of a legislative novel, the presumption of good faith of a participant in criminal proceedings is formulated.Presumption of good faith:Everyone is considered conscientious and reasonable, until otherwise is proved in the manner prescribed by this Code and established by a judicial act that has entered into legal force.Everyone is not obliged to prove their conscientiousness and reasonableness. The burden of proving dishonesty, unreasonableness, abuse of right lies with the person who claims the opposite.All doubts about the good faith and reasonableness of a person that cannot be eliminated in the manner prescribed by this Code shall be interpreted in favor of the person whose reputation is questioned.Any procedural actions and (or) decisions of a state body or an official in the field of criminal proceedings, even based on assumptions, cannot be based solely on the fact that a person has a certain procedural status in a case under consideration or another.

Exploiting mixing regularization for truly unsupervised font synthesis

Creating a novel font set requires domain expertise and is a laborious and time-consuming process, particularly for languages with a large number of characters and complicated structures. Existing deep learning based methods consider font generation (FG) as an image-to-image translation problem, mostly in a supervised setting, either in the form of pair images (paired data) or font labels (character or style labels), which requires extensive effort and is expensive to collect. Additionally, these supervised counter parts lack generalization for extending to other text image-related tasks, such as word image generation and font attribute control at inference time. We found that these drawbacks are mainly due to the supervised setting adopted by these existing methods for font generation. In this paper, we tackle the FG problem in a truly unsupervised fashion, where a complete font set can be generated by training the generator such that adjacent styles are not correlated and projecting the input glyph image into its corresponding font style latent space. To accomplish this, we propose the Font Mixing Generative Adversarial Network (FM-GAN), which employs mixing regularization to supervise the generator to localize the font styles, and a projection encoder to project an arbitrary glyph image into its corresponding semantic space that is compatible with the generator. In the experiments, we demonstrated that our unsupervised model synthesizes font images that are comparable to supervised state-of-the-art FG baselines. Furthermore, FM-GAN can be directly applied to other text image related tasks, such as multi-lingual font style transfer, word image generation, and font attribute control.

Visualization Guides in the Understanding of Theoretical Material in Lectures

The aim of this study is to demonstrate how students' comprehension is enhanced during the evaluation of a German language lecture through the use of visualization techniques. The research methodology utilized a descriptive quantitative approach and involved several tools such as video visualization, material exercises, questionnaires, and written documentation. Videos were used to assess language proficiency in various domains, including productive, receptive, and literary. The information presented through visualization utilized a combination of colors, varying word sizes, images, graphics, and audio. The study's findings indicate that students were able to comprehend approximately one-third of the subject matter due to the clarity and appeal of the visualizations. The inclusion of examples and instruction within the content contributed to a better understanding of the subject.

Offline HWR Accuracy Enhancement with Image Enhancement and Deep Learning Techniques

Handwriting recognition (HWR) is the ability of a machine to recognize handwritten text present in images, scanned or photographed. While much of the work reported in this field deals with scanned images, recognizing handwritten content from camera-captured images is still a challenge. While researchers are focusing their efforts to improve on the recognition process of such characters, the current work focuses efforts outside the recognition process.This work employs image enhancement techniques, noise removal, and binarization with adaptive thresholding & segmentation, to increase HWR accuracy. Additionally, the work of employs deep learning-based language models, such as BERT & BART, to further enhance the accuracy of recognized text. It is observed that HWR accuracy increases when smaller images such as word images are used as input to the HWR engine. On the post-processing front, BART is demonstrated to be superior in enhancing the accuracy of HWR-recognized text documents in a comparative study presented in this work. An accuracy improvement of 20+% is achieved with the BART model enhanced with transfer learning through a localized dataset to handle HWR-specific error patterns.